[GRASS5] Platform for next generation UI

[Trevor Wiens]

Michael and others,

There are have a number of calls to talk about general design first.
This is reasonable, because until we have a concrete image of what we
want, we can't make other decisions. In general I think there are 3
critical decisions to be undertaken.

1. QGIS or not. Are we willing to work within an existing framework and
live with the limitations that this brings. If so, then we need to
decide what existing aspects of this project need enhancement to meet
our goals and what parts need to be started to that our GUI needs are
met. Going with QGIS is the most conservative choice. We might not get
our ideal, but we will have something usable.

2. Tcltk or not. Although deciding on a toolkit early is not ideal,
there has been expressed and exists a very real and practical concern
about being able to succeed. By building on an existing foundations
there is a definite advantage, even if much of the existing code were
to be orphaned, at least the question of  CLI/GUI interaction is mostly
resolved by taking this step. Choosing Tcltk enhancement and redesign
is slightly more daring than choosing to build with QGIS, but it is
still quite likely to succeed.

3. Other toolkits. This is the most daring choice and the one most
likely to fail to deliver anything, but offers the chance of getting
the ideal. In this camp, probably vanilla Qt is the best choice (IMHO)
as it is going to be around for a long time and is mature. GTK and wx
seem like fine tools, but the cross platform issues with GTK rule it
out and as far as I know Qt is more complete than either. Use of KDE is
unwise as it ignores the diversity of our target users.

Personally, although I'm not all that keen on Tcltk, the flexibility of
going with an interpreted front end appeals to me. On this front I see
PyQt as another option. Whatever choice is made I think the that a
discussion and decision about joining the QGIS GUI needs to be made. On
this front I think perhaps listing things we don't like or would want to
change about QGIS might be helpful. For me these would be. 

1. Everything is a single window. Yuck.
2. For label placement a separate map layout page sucks.
The original design of Atlas GIS was much better. The page was
fundamental so there were no surprises about what your labels looked
like at print time. For more details on this refer to my comments in
Michael's summaries.
3. For symbols I would like to be able to vary a single symbol based on
two variables. For example change the size based on population and the
colour based on air quality. In this way a single symbol provides
information on two pieces of information. AFAIK, this is another old
Atlas feature not available anywhere that was very useful and would be
very useful to others. This can't be done in QGIS right now.

Otherwise, I quite like QGIS and see great potential in developing a
GRASS GUI in this manner. Depending on how flexible the QGIS team is,
they may be willing to enable a user choice of interface style from a
single window or multiple windows. This would take some work, but
considerably less than starting from scratch. Similarly the cartography
part of QGIS could be rewritten to do it right IMVAO (In my very
arrogant opinion). Enhancing the symbols would be minor.

Those problems fixed, one would have an excellent tool for serious use
both for analysis (building on the existing GRASS plugin) and
cartography.

T
--
Trevor Wiens 
twiens at interbaun.com

The significant problems that we face cannot be solved at the same 
level of thinking we were at when we created them. 
(Albert Einstein)